conv_fused.py

import math
import torch
import torch.nn as nn
import torch.nn.intrinsic as nni
import torch.nn.qat as nnqat
import torch.nn.functional as F
from torch.nn import init
from torch.nn.modules.utils import _single, _pair
from torch.nn.parameter import Parameter

_BN_CLASS_MAP = {
    1: nn.BatchNorm1d,
    2: nn.BatchNorm2d,
    3: nn.BatchNorm3d,
}


class _ConvBnNd(nn.modules.conv._ConvNd, nni._FusedModule):

    _version = 2

    def __init__(self,
                 # ConvNd args
                 in_channels, out_channels, kernel_size, stride,
                 padding, dilation, transposed, output_padding,
                 groups,
                 bias,
                 padding_mode,
                 # BatchNormNd args
                 # num_features: out_channels
                 eps=1e-05, momentum=0.1,
                 # affine: True
                 # track_running_stats: True
                 # Args for this module
                 freeze_bn=False,
                 qconfig=None,
                 dim=2):
        nn.modules.conv._ConvNd.__init__(self, in_channels, out_channels, kernel_size,
                                         stride, padding, dilation, transposed,
                                         output_padding, groups, False, padding_mode)
        assert qconfig, 'qconfig must be provided for QAT module'
        self.qconfig = qconfig
        self.freeze_bn = freeze_bn if self.training else True
        self.bn = _BN_CLASS_MAP[dim](out_channels, eps, momentum, True, True)
        self.weight_fake_quant = self.qconfig.weight()
        if bias:
            self.bias = Parameter(torch.Tensor(out_channels))
        else:
            self.register_parameter('bias', None)
        self.reset_bn_parameters()

        # this needs to be called after reset_bn_parameters,
        # as they modify the same state
        if self.training:
            if freeze_bn:
                self.freeze_bn_stats()
            else:
                self.update_bn_stats()
        else:
            self.freeze_bn_stats()

    def reset_running_stats(self):
        self.bn.reset_running_stats()

    def reset_bn_parameters(self):
        self.bn.reset_running_stats()
        init.uniform_(self.bn.weight)
        init.zeros_(self.bn.bias)
        # note: below is actully for conv, not BN
        if self.bias is not None:
            fan_in, _ = init._calculate_fan_in_and_fan_out(self.weight)
            bound = 1 / math.sqrt(fan_in)
            init.uniform_(self.bias, -bound, bound)

    def reset_parameters(self):
        super(_ConvBnNd, self).reset_parameters()

    def update_bn_stats(self):
        self.freeze_bn = False
        self.bn.training = True
        return self

    def freeze_bn_stats(self):
        self.freeze_bn = True
        self.bn.training = False
        return self

    def _forward(self, input):
        running_std = torch.sqrt(self.bn.running_var + self.bn.eps)
        scale_factor = self.bn.weight / running_std
        weight_shape = [1] * len(self.weight.shape)
        weight_shape[0] = -1
        bias_shape = [1] * len(self.weight.shape)
        bias_shape[1] = -1
        scaled_weight = self.weight_fake_quant(self.weight * scale_factor.reshape(weight_shape))
        # using zero bias here since the bias for original conv
        # will be added later
        if self.bias is not None:
            zero_bias = torch.zeros_like(self.bias)
        else:
            zero_bias = torch.zeros(self.out_channels, device=scaled_weight.device)
        conv = self._conv_forward(input, scaled_weight, zero_bias)
        conv_orig = conv / scale_factor.reshape(bias_shape)
        if self.bias is not None:
            conv_orig = conv_orig + self.bias.reshape(bias_shape)
        conv = self.bn(conv_orig)
        return conv

    def extra_repr(self):
        # TODO(jerryzh): extend
        return super(_ConvBnNd, self).extra_repr()

    def forward(self, input):
        return self._forward(input)

    def train(self, mode=True):
        """
        Batchnorm's training behavior is using the self.training flag. Prevent
        changing it if BN is frozen. This makes sure that calling `model.train()`
        on a model with a frozen BN will behave properly.
        """
        self.training = mode
        if not self.freeze_bn:
            for module in self.children():
                module.train(mode)
        return self

    # ===== Serialization version history =====
    #
    # Version 1/None
    #   self
    #   |--- weight : Tensor
    #   |--- bias : Tensor
    #   |--- gamma : Tensor
    #   |--- beta : Tensor
    #   |--- running_mean : Tensor
    #   |--- running_var : Tensor
    #   |--- num_batches_tracked : Tensor
    #
    # Version 2
    #   self
    #   |--- weight : Tensor
    #   |--- bias : Tensor
    #   |--- bn : Module
    #        |--- weight : Tensor (moved from v1.self.gamma)
    #        |--- bias : Tensor (moved from v1.self.beta)
    #        |--- running_mean : Tensor (moved from v1.self.running_mean)
    #        |--- running_var : Tensor (moved from v1.self.running_var)
    #        |--- num_batches_tracked : Tensor (moved from v1.self.num_batches_tracked)
    def _load_from_state_dict(self, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs):
        version = local_metadata.get('version', None)
        if version is None or version == 1:
            # BN related parameters and buffers were moved into the BN module for v2
            v2_to_v1_names = {
                'bn.weight': 'gamma',
                'bn.bias': 'beta',
                'bn.running_mean': 'running_mean',
                'bn.running_var': 'running_var',
                'bn.num_batches_tracked': 'num_batches_tracked',
            }
            for v2_name, v1_name in v2_to_v1_names.items():
                if prefix + v1_name in state_dict:
                    state_dict[prefix + v2_name] = state_dict[prefix + v1_name]
                    state_dict.pop(prefix + v1_name)
                elif prefix + v2_name in state_dict:
                    # there was a brief period where forward compatibility
                    # for this module was broken (between
                    # https://github.com/pytorch/pytorch/pull/38478
                    # and https://github.com/pytorch/pytorch/pull/38820)
                    # and modules emitted the v2 state_dict format while
                    # specifying that version == 1. This patches the forward
                    # compatibility issue by allowing the v2 style entries to
                    # be used.
                    pass
                elif strict:
                    missing_keys.append(prefix + v2_name)

        super(_ConvBnNd, self)._load_from_state_dict(
            state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs)

    @classmethod
    def from_float(cls, mod):
        r"""Create a qat module from a float module or qparams_dict

            Args: `mod` a float module, either produced by torch.quantization utilities
            or directly from user
        """
        assert type(mod) == cls._FLOAT_MODULE, 'qat.' + cls.__name__ + '.from_float only works for ' + \
            cls._FLOAT_MODULE.__name__
        assert hasattr(mod, 'qconfig'), 'Input float module must have qconfig defined'
        assert mod.qconfig, 'Input float module must have a valid qconfig'
        qconfig = mod.qconfig
        conv, bn = mod[0], mod[1]
        qat_convbn = cls(conv.in_channels, conv.out_channels, conv.kernel_size,
                         conv.stride, conv.padding, conv.dilation,
                         conv.groups, conv.bias is not None,
                         conv.padding_mode,
                         bn.eps, bn.momentum,
                         False,
                         qconfig)
        qat_convbn.weight = conv.weight
        qat_convbn.bias = conv.bias
        qat_convbn.bn.weight = bn.weight
        qat_convbn.bn.bias = bn.bias
        qat_convbn.bn.running_mean = bn.running_mean
        qat_convbn.bn.running_var = bn.running_var
        qat_convbn.bn.num_batches_tracked = bn.num_batches_tracked
        return qat_convbn

class ConvBn1d(_ConvBnNd, nn.Conv1d):
    r"""
    A ConvBn1d module is a module fused from Conv1d and BatchNorm1d,
    attached with FakeQuantize modules for weight,
    used in quantization aware training.

    We combined the interface of :class:`torch.nn.Conv1d` and
    :class:`torch.nn.BatchNorm1d`.

    Similar to :class:`torch.nn.Conv1d`, with FakeQuantize modules initialized
    to default.

    Attributes:
        freeze_bn:
        weight_fake_quant: fake quant module for weight

    """
    _FLOAT_MODULE = nni.ConvBn1d

    def __init__(self,
                 # Conv1d args
                 in_channels, out_channels, kernel_size, stride=1,
                 padding=0, dilation=1, groups=1,
                 bias=None,
                 padding_mode='zeros',
                 # BatchNorm1d args
                 # num_features: out_channels
                 eps=1e-05, momentum=0.1,
                 # affine: True
                 # track_running_stats: True
                 # Args for this module
                 freeze_bn=False,
                 qconfig=None):
        kernel_size = _single(kernel_size)
        stride = _single(stride)
        padding = _single(padding)
        dilation = _single(dilation)
        _ConvBnNd.__init__(self, in_channels, out_channels, kernel_size, stride,
                           padding, dilation, False, _single(0), groups, bias, padding_mode,
                           eps, momentum, freeze_bn, qconfig, dim=1)

class ConvBnReLU1d(ConvBn1d):
    r"""
    A ConvBnReLU1d module is a module fused from Conv1d, BatchNorm1d and ReLU,
    attached with FakeQuantize modules for weight,
    used in quantization aware training.

    We combined the interface of :class:`torch.nn.Conv1d` and
    :class:`torch.nn.BatchNorm1d` and :class:`torch.nn.ReLU`.

    Similar to `torch.nn.Conv1d`, with FakeQuantize modules initialized to
    default.

    Attributes:
        weight_fake_quant: fake quant module for weight

    """
    _FLOAT_MODULE = nni.ConvBnReLU1d

    def __init__(self,
                 # Conv1d args
                 in_channels, out_channels, kernel_size, stride=1,
                 padding=0, dilation=1, groups=1,
                 bias=None,
                 padding_mode='zeros',
                 # BatchNorm1d args
                 # num_features: out_channels
                 eps=1e-05, momentum=0.1,
                 # affine: True
                 # track_running_stats: True
                 # Args for this module
                 freeze_bn=False,
                 qconfig=None):
        super().__init__(in_channels, out_channels, kernel_size, stride,
                         padding, dilation, groups, bias,
                         padding_mode, eps, momentum,
                         freeze_bn,
                         qconfig)

    def forward(self, input):
        return F.relu(ConvBn1d._forward(self, input))

    @classmethod
    def from_float(cls, mod):
        return super(ConvBnReLU1d, cls).from_float(mod)

class ConvBn2d(_ConvBnNd, nn.Conv2d):
    r"""
    A ConvBn2d module is a module fused from Conv2d and BatchNorm2d,
    attached with FakeQuantize modules for weight,
    used in quantization aware training.

    We combined the interface of :class:`torch.nn.Conv2d` and
    :class:`torch.nn.BatchNorm2d`.

    Similar to :class:`torch.nn.Conv2d`, with FakeQuantize modules initialized
    to default.

    Attributes:
        freeze_bn:
        weight_fake_quant: fake quant module for weight

    """
    _FLOAT_MODULE = nni.ConvBn2d

    def __init__(self,
                 # ConvNd args
                 in_channels, out_channels, kernel_size, stride=1,
                 padding=0, dilation=1, groups=1,
                 bias=None,
                 padding_mode='zeros',
                 # BatchNorm2d args
                 # num_features: out_channels
                 eps=1e-05, momentum=0.1,
                 # affine: True
                 # track_running_stats: True
                 # Args for this module
                 freeze_bn=False,
                 qconfig=None):
        kernel_size = _pair(kernel_size)
        stride = _pair(stride)
        padding = _pair(padding)
        dilation = _pair(dilation)
        _ConvBnNd.__init__(self, in_channels, out_channels, kernel_size, stride,
                           padding, dilation, False, _pair(0), groups, bias, padding_mode,
                           eps, momentum, freeze_bn, qconfig, dim=2)

class ConvBnReLU2d(ConvBn2d):
    r"""
    A ConvBnReLU2d module is a module fused from Conv2d, BatchNorm2d and ReLU,
    attached with FakeQuantize modules for weight,
    used in quantization aware training.

    We combined the interface of :class:`torch.nn.Conv2d` and
    :class:`torch.nn.BatchNorm2d` and :class:`torch.nn.ReLU`.

    Similar to `torch.nn.Conv2d`, with FakeQuantize modules initialized to
    default.

    Attributes:
        weight_fake_quant: fake quant module for weight

    """
    _FLOAT_MODULE = nni.ConvBnReLU2d

    def __init__(self,
                 # Conv2d args
                 in_channels, out_channels, kernel_size, stride=1,
                 padding=0, dilation=1, groups=1,
                 bias=None,
                 padding_mode='zeros',
                 # BatchNorm2d args
                 # num_features: out_channels
                 eps=1e-05, momentum=0.1,
                 # affine: True
                 # track_running_stats: True
                 # Args for this module
                 freeze_bn=False,
                 qconfig=None):
        super(ConvBnReLU2d, self).__init__(in_channels, out_channels, kernel_size, stride,
                                           padding, dilation, groups, bias,
                                           padding_mode, eps, momentum,
                                           freeze_bn,
                                           qconfig)

    def forward(self, input):
        return F.relu(ConvBn2d._forward(self, input))

    @classmethod
    def from_float(cls, mod):
        return super(ConvBnReLU2d, cls).from_float(mod)

class ConvReLU2d(nnqat.Conv2d, nni._FusedModule):
    r"""A ConvReLU2d module is a fused module of Conv2d and ReLU, attached with
    FakeQuantize modules for weight for
    quantization aware training.

    We combined the interface of :class:`~torch.nn.Conv2d` and
    :class:`~torch.nn.BatchNorm2d`.

    Attributes:
        weight_fake_quant: fake quant module for weight

    """
    _FLOAT_MODULE = nni.ConvReLU2d

    def __init__(self, in_channels, out_channels, kernel_size, stride=1,
                 padding=0, dilation=1, groups=1,
                 bias=True, padding_mode='zeros',
                 qconfig=None):
        super(ConvReLU2d, self).__init__(in_channels, out_channels, kernel_size,
                                         stride=stride, padding=padding, dilation=dilation,
                                         groups=groups, bias=bias, padding_mode=padding_mode,
                                         qconfig=qconfig)
        assert qconfig, 'qconfig must be provided for QAT module'
        self.qconfig = qconfig
        self.weight_fake_quant = self.qconfig.weight()

    def forward(self, input):
        return F.relu(
            self._conv_forward(input, self.weight_fake_quant(self.weight), self.bias))

    @classmethod
    def from_float(cls, mod):
        return super(ConvReLU2d, cls).from_float(mod)

def update_bn_stats(mod):
    if type(mod) in set([ConvBnReLU1d, ConvBnReLU2d, ConvBn1d, ConvBn2d]):
        mod.update_bn_stats()

def freeze_bn_stats(mod):
    if type(mod) in set([ConvBnReLU1d, ConvBnReLU2d, ConvBn1d, ConvBn2d]):
        mod.freeze_bn_stats()